Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Our models should be robust enough to translate a calendar #600

Closed
Tracked by #238
gregtatum opened this issue May 15, 2024 · 3 comments
Closed
Tracked by #238

Our models should be robust enough to translate a calendar #600

gregtatum opened this issue May 15, 2024 · 3 comments
Labels
quality Improving robustness and translation quality

Comments

@gregtatum
Copy link
Member

The models tend to break down when presented with single numbers, or lists of numbers.

For example de-en translates

1
2
3
4
5
6
7
8
9

as

1
2
3
4
5 5
6 6
7
8
9 9

While en-ca (which did not use opus cleaner) succeeds.

Here is an example of what calendars look like when the model can't handle this.

screenshot of a calendar widget with normal numbers

![screenshot of a calendar widget with repeated numbers]https://github.com/mozilla/firefox-translations-training/assets/1588648/28d570fe-cc11-42f3-b7ed-d314575bc5d8)

This page is pretty useful as an in-browser test: https://www.timeanddate.com/calendar/

@gregtatum gregtatum added the quality Improving robustness and translation quality label May 15, 2024
@gregtatum
Copy link
Member Author

This could be helped by num_mismatch in OpusCleaner, and probably #215.

@marco-c
Copy link
Collaborator

marco-c commented May 15, 2024

Sometimes I wonder if it makes sense to implement some hardcoded rules.

@eu9ene
Copy link
Collaborator

eu9ene commented Jul 12, 2024

I see that the new models (lt-en, sl-en, sr-en, lv-en) translate numbers correctly:

Screenshot 2024-07-12 at 3 58 20 PM

Closing. Retraining existing models to use the latest improvements is a separate issue.

@eu9ene eu9ene closed this as completed Jul 12, 2024
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Labels
quality Improving robustness and translation quality
Projects
None yet
Development

No branches or pull requests

3 participants